From Feb 2011 to May 2012, children aged 6 to 59 months presenting with diarrhea to the Hospital Padre Alberto Buffoni (HPAB) in Quininde, Ecuador and surrounding family health clinics were considered for enrolment. Diarrhea was defined as three or more liquid or semi-liquid stools in 24 hours, with duration less than 14 days before consultation with the clinic/hospital. When a child presented with diarrhea, a nurse collected a fecal specimen and clinical data. Fecal specimens were tested for presence of rotavirus antigen by enzyme immunoassay (EIA). Children testing positive for rotavirus by EIA were defined as cases. Two comparison groups were recruited: (1) children who presented with diarrhea but tested negative for rotavirus by EIA (henceforth referred to as diarrhea controls) and healthy children without diarrhea, vomiting or hospitalization for any cause 14 days prior to recruitment and members of an ongoing birth cohort study who presented to the HPAB for routine follow-ups [10] (henceforth referred to as healthy controls). Both groups of controls were matched for age (+/−6 months for cases under age 3 and +/−1 year for cases aged 3 to 4 years).

After cases of rotavirus were detected, fecal samples were requested from all adult and child household members of the case during a household visit. Households were visited twice. At the first visit, the head of the household was informed about the household study, given consent procedures, and specimen collection pots were left for each household member. Henceforth, we refer to the three household populations as case households, diarrhea control households, and healthy control households, based on the disease and infections status of the index case. Specimen collection was requested as the date of onset of diarrhea in the index child (case or control) +7 days or as near as possible (5 to 9 days after onset). An adult representative of the household was interviewed for information on demographics, date of specimen collection and presence of diarrhea in the 10 days before and 10 days after the onset of diarrhea in the index case. For healthy control households, specimens were taken within one day of the household visit and information was collected on any symptoms in the 10 days before the index date of recruitment.

This study and consent procedures were reviewed and approved by the Bioethics Committee of the Universidad de San Francisco de Quito. All recruited household members were explained the premise of the project and were asked to provide written informed consent prior to enrollment in the study. A parent or guardian was asked to provide written consent on behalf of minors in the household.

Specimens taken following episodes of diarrhea in index cases were tested for rotavirus using the Prospect™ Rotavirus Test (Oxoid Diagnostics Ltd, United Kingdom). This test was used to initially determine if the diarrhea presentations in the clinic/hospital were cases (rotavirus-positive) or diarrhea controls (rotavirus negative). The Prospect™ kit and other validated enzyme immunoassays are calibrated to detect rotavirus at levels considered to cause disease. The majority of rotavirus infections in household members were expected to be asymptomatic. In order to detect asymptomatic infections, real time reverse transcription and quantitative polymerase chain reaction (RT-qPCR) was used, due to its much greater analytic sensitivity [11]–[13]. Note that stools of all symptomatic index children (cases and diarrhea controls) were tested by RT-qPCR (following EIA), while household contact were only tested by RT-qPCR.

Remaining specimens were stored at −20°C for conventional RT-PCR and genotyping. Genotyping by RT-PCR was performed on positive samples to confirm transmission within a household (i.e. to confirm that secondary cases are infected with the same type as the primary cases in the household).

For EIA or RT-qPCR confirmed positive samples, total RNA was extracted by using the MagMax Viral RNA Isolation kit (Life Technologies, New York, NY) on the automated KingFisher extraction platform (Thermo Electron Corporation, Vantaa, Finland).

G-type (VP7) and P-type (VP4) genotyping were carried out following previously described 2-step amplification methods [14], [15] with modification on a GeneAMP PCR System 9700 thermocycler (Applied Biosystems, Foster City, CA). VP7 and VP4 were amplified by RT-PCR using consensus primers 9Con1-L/VP7R [14] and Con3/Con2, [15] respectively. The RT-PCR step was modified to use the One-Step RT-PCR kit (Qiagen, Inc., Valencia, CA) on a GeneAMP PCR System 9700 thermocycler (Applied Biosystems, Foster City, CA). In this procedure, the extracted dsRNA was denatured at 97°C for 5 min and RT-PCR was carried out using a One Step RT-PCR kit (Qiagen, Inc., Valencia, CA) according to manufacturer’s instructions. Reverse transcription (RT) of each gene was carried out for 30 min at 42°C, followed by 15 min at 95°C to inactivate the reverse transcriptase and activate the Taq polymerase. The cDNA was then subjected to 35 cycles of PCR in a GeneAmp PCR System 9700 thermal cycler (Applied Biosystems, Inc., Foster City, CA) using the following conditions: 30 sec at 94°C; 30 sec at 42°C; 45 sec at 72°C, followed by a 7 min final extension at 72°C. Genotyping PCR was performed as described previously. [14], [15] G-typing used primer 9Con1-L in combination with primers 9T1–1, 9T1-Dg, 9T-2, 9T-3P, 9T-4, and 9T-9B [14] and P-typing used primer con3 in combination with primers 1T-1, 1T1-VN, 2T-1, 3T-1, 4T-1, 5T-1, and 1T1-Wa [15]. Genotyping reactions were analyzed by electrophoresis on a 3% agarose gel using a 2∶1 ratio of NuSieve GTG: SeaPlaque (FMC Bioproducts, Rockland, ME). The G- and P-types obtained were classified according to the system described by Estes and Cohen [16]. G1 viruses were examined by specific primers to determine if they were vaccine (Rotarix) strains.

A selected number of isolates from the same household with the same genotyping results were subjected to sequencing for confirmation. The same consensus primer pair con3/con2 for VP4 gene and 9Con1-L/VP7R for VP7 gene were used to generate amplicons for the sequence reaction. Analysis of RT-PCR reactions by gel electrophoresis, amplicon purification, and DNA sequencing was carried out as described previously [17]. Forward and reverse sequences were assembled using Sequencher software versions 4.8 (Gene Codes Corporation, Inc, Ann Arbor, MI). The consensus sequences obtained were compared with existing rotavirus sequences in the GenBank database using the BLASTN program at the National Center for Biotechnology Information website (available at: http://www.ncbi.nlm.gov/BLAST/).

Infection amongst household contacts was defined as rotavirus detected by RT-qPCR in stool at the time of specimen collection; disease was defined as diarrheal disease symptoms within 10 days of onset in the index case. We also report disease attack rates based on diarrheal disease symptoms and detection of rotavirus by RT-qPCR, since it is possible that some of the PCR-negative individuals were symptomatic due to something other than rotavirus. All new viral sequence data has been deposited in GenBank; strain names and accession numbers can be found in File S1. Accession numbers for submitted strains.

We investigated potential risk factors for transmissibility (characteristics of cases) and susceptibility (characteristics of household contacts). Only household contacts of rotavirus index cases (n = 137) were included in the statistical analysis of transmission. We calculated the infection and disease attack rates and fit logistic regression models to estimate the odds ratio separately for each outcome. Initially, we attempted to fit random effects logistic regression models in order to explicitly account for correlation in outcomes with households. However, a number of bivariate models could not achieve convergence, so instead of fitting a random effects model, robust standard errors were calculated, treating each household as a cluster. First, bivariate models were fitted for each potential transmissibility and susceptibility factor. Separate models were fitted, with infection and disease as separate binary outcomes.

Then, we attempted to control for confounding by developing multivariable regression models. First, two separate were constructed by inclusion of all variables with p<0.20 in univariable analysis. The first model included only case (transmissibility) variables; the second included only contact (susceptibility) variables Then, using the same criteria, final models were fit with both transmissibility and susceptibility variables. Only one of a set of highly collinear variables (e.g. vomiting duration and severity score) was retained in multivariable models. Analyses were conducted in Stata 12.0 (STATA Corp, College Station, Tx).

Infection was detected by RT-qPCR in 55% (76/137) of case household contacts, 7.6% (10/130) of diarrhea control contacts and 2.2% (3/137) healthy control contacts (Figure 2).

The effects of index case characteristics (i.e. transmissibility) (i.e. index case characteristics) and contact characteristics (i.e. susceptibility) on infection attack rates are shown in Table 1. Regarding transmissibility, infection attack rates were non-significantly higher from younger children (67% from children 6 to 17 months compared to 53% from children ≥18 months); those who had vomiting (63% compared to 37%; OR = 2.87, p = 0.099) and those with Vesikari scores ≥10 (66% compared to 43%; OR = 2.58, p = 0.086), although these differences were of borderline statistical significance (Table 1 and Figure 3). Children with higher cycle threshold (Ct) values (indicative of lower viral load) were more likely to transmit though not at a level of statistical significance (62% with Ct <15; 40% with Ct was ≥15; OR = 0.41, p = 0.177). We did not find evidence of reduced transmissibility associated with a history of rotavirus vaccine, though study power was limited due to high vaccine coverage; 85% and 87% of index cases had one or two doses of vaccine, respectively.

Regarding susceptibility, infection attack rates were higher amongst siblings (70%; OR = 3.83, p = 0.015) and parents (56%; OR = 2.09, p = 0.17) compared to other household members (e.g. aunts/uncles/grandparents; 38%).

The higher risk of infection in siblings remained significant when controlling for other case and contact variables in multivariable models; greater transmissibility associated with vomiting was of borderline significance when accounting for other index case characteristics (OR = 2.50, p = 0.079) as well as in the full models accounting for contact characteristics (OR = 2.42, p = 0.135).

There were no episodes of diarrhea reported in the healthy control households in the 10 days before enrolment. Five diarrhea episodes were reported in diarrhea control households (attack rate = 4%), all of which occurred within 3 days of onset in the index case. Twenty acute gastroenteritis (AGE) episodes were reported in case households (attack rate = 15%), six of which had times of onset 5 to 1 days prior to onset in the index case and 14 with onset 1 to 5 days after onset in the index case (Figure 2; Table 2). Only these 14 cases were considered as possible secondary episodes for subsequent analyses. 8 of the 14 possible secondary episodes were positive for rotavirus by RT-qPCR. All of the cases with onset prior to the index case were amongst older individuals (aged 5 to 37 years).

The effects of index case characteristics (i.e. transmissibility) (i.e. index case characteristics) and contact characteristics (i.e susceptibility) on disease attack rates are shown in Table 2. Regarding transmissibility, disease attack rates were higher from children who were younger (48% from children <18 months compared to 2% from children ≥18 months) or had multiple episodes of vomiting (23% compared to 3% from children with 1 or no episodes of vomiting; OR = 11.6, p = 0.031) (Figure 3).

Regarding susceptibility to disease, attack rates were higher amongst contacts aged below 10 years (31%; OR = 4.44, p = 0.001 compared to contact aged 30 years or older) and those who shared a room with the index case (17% compared with 2%; OR = 11.5, p = 0.056).

There were only two instances where a household contact was younger than the index case. Both of these children were infected and one was symptomatic.

In the final disease model that controlled for case and contact characteristics, children <18 months were more infectious (OR = 92, p<0.001), multiple vomiting episodes were associated with high transmission risk (OR = 3.63; p = 0.02) and children aged <10 years were most susceptible to disease (OR = 19.7, p = 0.01).

Rotavirus was detected by RT-qPCR in stool from 33% (13/40) of diarrhea controls and 12% (5/40) healthy controls, all of whom were negative by EIA. Infection attack rates were 16% (7 out of 43 household members) in the diarrhea control households and 0% (out of 20 household members) in infected healthy control households. None of the household members reported diarrhea in the 10 days prior to the onset in the index healthy control, but there were 3 symptomatic cases after onset in the index diarrhea control. The Ct value of the diarrhea control was >34 in these three instances (Figure 4). Further, rotavirus was not detected in any of these three symptomatic contacts cases, suggesting that the cause of diarrhea in both the index diarrhea control and the household member was not rotavirus.

We were able to determine genotypes from 36 of the 39 index cases. G2P [4] genotype was detected from 16 (44%) of these, and G2P [8] and G4P [4] was detected in one sample each. G9P was detected from 14 specimens, 12 of which had a P-type P [8], one with P [4] and one could not be P-typed. Four specimens were G3, two typed with P [8], and one each with P [4] and P [6].

Genotyping data were available from 52 contacts in 22 case households where the genotype of the index case was also determined (Figure 5). 38 (73%) of the 52 contacts shared a G-type or P-type with the index case. However, in only seven of the 22 households (32%) was a single genotype found amongst the index case and all household contacts. In some households, there was a single household contact with a different genotype than the index case. For example, in one household the index case was G9P [8], along with 2 household contacts, while another household contact was a G1. Still in other households, there was clear evidence of circulation of multiple viruses. For example, in one household, G2P [4] was detected in the index case, G9P [8] in one contact with both viruses in another household member. In addition, P [8] was detected in the index case and G [2] was detected in the G9P [8]-infected contact, suggesting that all three individuals were infected with both viruses.

A total of 10 G1 viruses were detected (1 healthy control, 1 diarrhea control, 2 household contact of diarrhea controls and 7 household contacts of cases). None were Rotarix vaccine strain.